Research And Implementation Of Sdn Based Flexible Deployment System Of Transition Technology

In recent years,with the progress of hardware computing power and the development of deep learning technology,the algorithm based on deep learning has achieved good results in face detection.However,when the face is occluded,the current algorithm will cause the problem of false detection and missed detection of the face,which affects its accuracy.This paper studies face detection and occlusion recognition under partial occlusion.Firstly,face occlusion recognition algorithm is proposed to meet the requirements of unocclusion and clear image.Experiments show that the recognition using classification algorithm as occlusion type and occlusion part is faster than the occlusion recognition algorithm based on key point detection,and it can more accurately identify the occlusion parts and occlusion categories of occluded faces,such as whether the left eye,right eye,nose and other parts are occluded,whether the side face,whether fuzzy,whether wearing glasses and glasses,whether wearing masks,etc.Secondly,in order to meet the requirements of automatic clipping of documents and solve the quality problem of occlusion recognition face data,Based on the existing network structure,the face detection algorithm of masking is studied and a dual threshold Soft-NMS algorithm based on improved non-maximum suppression(NMS)is proposed.Experimental results show that the proposed algorithm can effectively improve the detection accuracy of occluded face.Finally,this paper designs and implements a graduation certificate collection system,and tests the system.